Optimizing NASA’s High-Performance Computing: Hardware Improvements and Storage Streamlining to Reduce “Time-To-Science” on Physical Science Projects and Simulations

Poster Session
STARR
Poster Number
B-09
Project Author(s)
Alexander Mote
Institution
Oregon State University
Project Description

High-performance computing has quickly become an integral part of scientific advancement, and a great deal of work has gone into making deep learning software easier to both create and deploy; however, many scientific projects have been bottlenecked by hardware limitations, rather than software ones. This research examines those limitations and investigates what changes can be made to minimize them. FPGAs (Field Programmable Gate Arrays) prove to be a promising alternative to GPUs (Graphical Processing Units) in both power consumption and interface flexibility, and storage optimization allows for these accelerators to access data more quickly and without the need for a CPU (Central Processing Unit) interface. Further research should develop models for optimizing FPGA integration, and test physical science algorithms on different processor and storage models to find optimal configurations.

 

Remote video URL